U.S. patent application number 10/780430 was filed with the patent office on 2005-08-18 for reference phase and amplitude estimation for coherent optical receiver.
Invention is credited to Feced, Ricardo, Richard, Epworth, Rickard, Robin.
Application Number | 20050180760 10/780430 |
Document ID | / |
Family ID | 34838590 |
Filed Date | 2005-08-18 |
United States Patent
Application |
20050180760 |
Kind Code |
A1 |
Feced, Ricardo ; et
al. |
August 18, 2005 |
Reference phase and amplitude estimation for coherent optical
receiver
Abstract
An optical receiver demodulates optical orthogonal frequency
division multiplexed signals and generates a number of subcarrier
reference signals, each for demodulating a frequency channel of the
frequency division multiplexed signals. It compensates for
degradations in the generated reference signals by averaging a
number of estimates derived from different inputs to make the
references more resilient to degradations. It can encompass time
averaging to compensate for amplification noise, and frequency
averaging of phase drift estimation to compensate for phase drift
caused by reduced source coherence. It can enable longer system
reach and/or increased optical power margins by means of better
system resilience to amplification noise and reduced source
coherence. The bit error rate can be reduced, and/or the capacity
can be increased by increasing bit rate or introducing more
frequency channels.
Inventors: |
Feced, Ricardo;
(Sawbridgeworth, GB) ; Rickard, Robin;
(Spellbrook, GB) ; Richard, Epworth;
(Sawbridgeworth, GB) |
Correspondence
Address: |
BARNES & THORNBURG
P.O. BOX 2786
CHICAGO
IL
60690-2786
US
|
Family ID: |
34838590 |
Appl. No.: |
10/780430 |
Filed: |
February 17, 2004 |
Current U.S.
Class: |
398/183 |
Current CPC
Class: |
H04B 10/2575 20130101;
H04B 10/548 20130101; H04B 10/60 20130101 |
Class at
Publication: |
398/183 |
International
Class: |
H04B 010/04 |
Claims
1. An optical receiver arranged to receive and demodulate optical
orthogonal frequency division multiplexed signals, and having a
subcarrier reference generator arranged to generate a number of
subcarrier reference signals, each for use in demodulating a
different one of a number of frequency channels of the frequency
division multiplexed signals, the subcarrier reference generator
further being arranged to compensate for degradations in the
generated reference signals by averaging a number of estimates
derived from different inputs.
2. The optical receiver of claim 1, the generator being arranged to
compensate for phase drift by determining estimates at a number of
different frequencies and averaging these estimates.
3. The optical receiver of claim 1, the generator being arranged to
compensate for noise by determining estimates for a given one of
the reference signals at a number of different times and averaging
these estimates.
4. The optical receiver of claim 1, the generator being arranged to
compensate for phase drift by determining estimates at a number of
different frequencies and averaging these estimates, and to
compensate for noise by determining estimates for a given one of
the reference signals at a number of different times and averaging
these estimates.
5. The optical receiver of claim 1, arranged to demodulate
differentially coded optical orthogonal frequency division
multiplexed signals, the generator being arranged to operate
without using a transmitted pilot tone.
6. The optical receiver of claim 1, arranged to demodulate non
differentially coded optical orthogonal frequency division
multiplexed signals.
7. The optical receiver of claim 1, the generator being arranged to
generate the estimated reference signals by stripping detected data
from a received signal for each channel.
8. The optical receiver of claim 1, the generator being arranged to
generate an estimated phase drift by estimating a rotation of a
constellation of points represented in a complex frequency
domain.
9. The optical receiver of claim 8, being arranged to receive
signals carrying data values which have been encoded before
transmission by a rotation in a complex frequency domain, relative
to other data values and the receiver having a decoder for decoding
by using the other data values to determine an amount of rotation
of the received data values, needed to decode the data values.
10. The optical receiver of claim 1, having a coherent optical
domain receiving section, and an electrical demodulation section
arranged to carry out a Fast Fourier Transform, and to use the
subcarrier reference signals for detection of data in the fourier
transformed signals.
11. A demodulator for a receiver for orthogonal frequency division
multiplexed signals, and having a subcarrier reference generator
arranged to generate a number of subcarrier reference signals, each
for use in demodulating a different one of a number of frequency
channels of the frequency division multiplexed signals, the
subcarrier reference generator further being arranged to compensate
for degradations in the generated reference signals by averaging a
number of estimates derived from different inputs.
12. The demodulator of claim 11, the generator being arranged to
compensate for phase drift by determining estimates at a number of
different frequencies and averaging these estimates, and to
compensate for noise by determining estimates for a given one of
the reference signals at a number of different times and averaging
these estimates.
13. The demodulator of claim 11, arranged to demodulate
differentially coded optical orthogonal frequency division
multiplexed signals, the generator being arranged to operate
without using a transmitted pilot tone.
14. The demodulator of claim 11 in the form of software.
15. A subcarrier reference generator for a demodulator for
orthogonal frequency division multiplexed signals, the generator
being arranged to generate a number of subcarrier reference
signals, each for use in demodulating a different one of a number
of frequency channels of the frequency division multiplexed
signals, the subcarrier reference generator further being arranged
to compensate for degradations in the generated reference signals
by averaging a number of estimates derived from different
inputs.
16. The subcarrier reference generator of claim 15, arranged to
demodulate differentially coded optical orthogonal frequency
division multiplexed signals, and being arranged to operate without
using a transmitted pilot tone.
17. A subcarrier reference generator for generating subcarrier
reference signals for use in demodulating optically transmitted
orthogonal frequency division multiplexed signals, by generating a
number of subcarrier reference signals, each for use in
demodulating a different one of a number of frequency channels of
the frequency division multiplexed signals, and comprising means
for compensating for degradations in the generated reference
signals by averaging a number of estimates derived from different
inputs.
18. An optical receiver arranged to receive and demodulate optical
orthogonal frequency division multiplexed signals carrying QAM data
values encoded by a mapping in a complex frequency domain according
to corresponding other data values, the receiver having a
subcarrier reference generator arranged to generate a number of
subcarrier reference signals, a demodulator for demodulating the
optical orthogonal frequency division multiplexed signals using
each of the subcarrier reference signals for a respective frequency
channel of the frequency division multiplexed signals, and a
decoder for decoding after the demodulating, by determining from
the corresponding other data values, an inverse mapping in the
complex frequency domain needed to decode the data values.
19. The optical receiver of claim 18, the mapping and inverse
mapping comprising a rotation.
20. The optical receiver of claim 19, the rotation being any of 0,
90, 180, or 270 degrees.
21. An optical transmitter arranged to transmit an optical
orthogonal frequency division multiplexed signal carrying QAM data
values, the transmitter having: an encoder for encoding the data
values by carrying out a mapping in a complex frequency domain
according to corresponding others of the data values, and a
modulator for modulating the encoded data values to form the
optical orthogonal frequency division multiplexed signal having a
number of frequency channels.
22. The optical transmitter of claim 21, the mapping comprising a
rotation.
23. A method of generating subcarrier reference signals for use in
demodulating optically transmitted orthogonal frequency division
multiplexed signals, by generating a number of subcarrier reference
signals, each for use in demodulating a different one of a number
of frequency channels of the frequency division multiplexed
signals, and by compensating for degradations in the generated
reference signals by averaging a number of estimates derived from
different inputs.
24. The method of claim 23, having the step of compensating for
phase drift by determining estimates at a number of different
frequencies and averaging these estimates.
25. A method of offering a data transmission service over an
optical link, the data being transmitted using optical orthogonal
frequency division multiplexed signals, and being received using an
optical receiver having a subcarrier reference generator arranged
to generate a number of subcarrier reference signals, each for use
in demodulating a different one of a number of frequency channels
of the frequency division multiplexed signals, the subcarrier
reference generator further being arranged to compensate for
degradations in the generated reference signals by averaging a
number of estimates derived from different inputs.
Description
RELATED APPLICATIONS
[0001] This invention relates to U.S. patent application entitled
"Optical Sub-carrier Multiplexed Transmission" (Nortel Networks
file 16231ID) Ser. No. 10/679,824 filed on 6 Oct. 2003 and U.S.
patent application entitled "MLSE across closely coupled optical
channels" (Nortel Networks file 15980ID) Ser. No. 10/626,384 filed
on 23 Jul. 2003, both incorporated herein by reference in their
entirety.
FIELD OF THE INVENTION
[0002] This invention relates to receivers, to demodulators, to
subcarrier reference generators, phase noise correctors, to
corresponding software, to corresponding methods and to methods of
offering a transmission service over such apparatus.
BACKGROUND TO THE INVENTION
[0003] Orthogonal Frequency Division Multiplexing (OFDM)
communication systems exhibit high resilience to linear distortion
impairments because the information is encoded in the frequency
domain. For this reason, the implementation of an optical OFDM
system has become an option actively investigated, as shown in the
above referenced pending U.S. patent application Ser. No.
10/679,824. However optical communication systems suffer more
severely from certain impairments than radio communication systems,
for which the OFDM technique was initially developed. The
successful use of OFDM in the optical arena will depend on how
satisfactorily it is possible to ameliorate the impact of these
impairments.
[0004] One of the main limitations of optical communication systems
is the optical noise generated in amplification stages which are
usually needed every 50-100 km. Consequently, it is of crucial
importance to establish the optimum demodulation technique in order
to minimise the impact of noise without adding any unnecessary
overheads.
[0005] Another limitation of coherent-type optical communication
systems is the reduced coherence of the laser optical sources in
comparison with their radiofrequency counterparts. It is also
necessary to establish adequate phase-noise reduction techniques to
be able to operate optical OFDM systems with realistic and
currently available optical sources.
[0006] Orthogonal Frequency Division Multiplexing (OFDM) systems
encode the information as a concatenation of blocks in the
frequency domain. An inverse fast Fourier transform converts the
information to the time domain before its transmission along the
channel. A fast Fourier transform at the receiver recovers the
original transmitted sequence.
[0007] The main advantage of OFDM is that if a guard-band with
cyclic prefix is included before transmission, then the received
sequence is the circular convolution of the input signal with the
channel response. This permits a very easy equalisation of linear
impairments, as circular convolutions become products in the
Fourier domain where the information is encoded. The only
limitation is that the time interval over which the channel
impulse-response-time extends should be smaller than the size of
the allocated guard-band.
[0008] It is also known to apply MLSE (Maximum Likelihood Sequence
Estimation) to multi-carrier radio receivers using joint
demodulation techniques to reduce co-channel interference (ICI). In
this joint demodulation technique, symbols of the desired signal
are decoded by an MLSE process at the same time as symbols of a
dominant interference source. The estimated contribution of the
interference source is subtracted out to decode the desired signal
and the estimated contribution of the desired signal is subtracted
out to decode the interference source. All possibilities of the
received contribution from the desired signal and the interference
source are tried, and a "score" (e.g., a viterbi decoding metric)
is kept for each.
[0009] However, ICI in radio transmissions is fundamentally
different to ICI in optical fibers and so radio techniques are not
generally appropriate for optical systems. One such difference is
that the transmission medium is dispersive in optical systems,
meaning that signals of different frequencies travel at different
velocities. Another fundamental difference is that the nonlinear
cross coupling between intensity and phase occurs along the fiber,
whereas in radio systems, there is no such distributed
nonlinearity. In contrast, nonlinearities in radio systems are
highly localized, being caused by a very few discrete nonlinear
elements. Each such localized nonlinearity is readily compensated
using a discrete compensating nonlinearity. This method is not
feasible in an optical fiber system, as the fiber nonlinearity
interacts with the dispersion in a complex distributed manner.
Hence most cross coupling or ICI in radio systems can be modeled
and pre or post compensated, whereas this is not effective for
optical ICI. Also, notably the bit rates in radio systems are lower
and so there is more time within the bit period, for complex
processing techniques. Hence where the reach performance is limited
by optical nonlinearity in the fiber, current practice is to
minimise nonlinear interaction between channels by spacing the
channels in frequency and using chromatic dispersion to minimise
the nonlinear interaction/cross coupling (by walk-off/dephasing).
Where the modulation format uses multiple-phases (such as QPSK), it
is currently necessary to reduce the operating power levels to
minimize the impairments due to nonlinear coupling between these
nominally orthogonal phase channels.
[0010] If OFDM is implemented in optical communication systems as
shown in the above referenced pending U.S. patent application Ser.
No. 10/679,824, the transmitted information can be encoded in
frequency by means of a given phase/amplitude modulation format
like QPSK, QAM 16, QAM 32 and similar formats. The polarisation
dimension can also be exploited by polarisation multiplexing. The
receiver employs a coherent-type detection scheme whereby the
in-phase and quadrature components of the signal are measured.
Schemes that detect either only one polarisation or two orthogonal
polarisations can be implemented, the second one being preferred to
enhance transmission capacity. Linear impairments, like chromatic
and polarisation mode dispersion, can be easily equalised once the
received sampled signal is converted back to the frequency domain
to extract the original data.
[0011] Standard OFDM for radio systems offers two approaches for
data demodulation, either in a pilot-tone based scheme or in a
differentially-detection scheme:
[0012] Pilot-tone scheme: reference pilot-tones with known data are
transmitted along with the desired information, and used in the
receiver to recover a reference signal for each frequency channel
(subcarrier) to demodulate the data. The reference signal is built
along time by an averaging process, so that it is virtually
noiseless, providing good resilience against additive white
Gaussian noise as it attains the coherent detection limit. However,
the transmission of pilot tones containing known information
constitutes an overhead that wastes part of the channel capacity,
and which is not necessary in the differentially-detection scheme.
Also the complexity of the receiver will be high.
[0013] Differential detection scheme: In this scheme the data is
demodulated by comparing consecutive sampled symbols corresponding
to the same frequency channel. There is a performance difference
between the two schemes in relation to their resilience to additive
white Gaussian noise (AWGN), and the possibility of phase error
correction. No overhead is required but it exhibits a
signal-to-noise ratio penalty with respect to pilot tones. No phase
error correction technique has been implemented in this scheme.
[0014] It is known from EBU technical review summer 1998, J Stott,
"The effects of Phase noise in COFDM", that there are two types of
phase noise, a common phase error common to all the channels and a
thermal noise-like part which is not pure phase noise. COFDM is
coded orthogonal frequency division multiplexing. It shows that a
pilot tone scheme can be implemented with phase error correction
circuitry in order to correct for common phase errors in a digital
video broadcast system.
[0015] In optical systems, AWGN introduced at each amplification
stage is usually one of the performance limiting factors. The
OFDM-pilot-tone based scheme exhibits the same Bit Error Rate (BER)
versus Optical Signal-to-Noise Ratio (OSNR) performance as its
equivalent coherently detected modulation formats. Analogously, the
OFDM-differentially-detecte- d scheme also exhibits the same
BER-OSNR performance as its equivalent differentially-detected
modulation format. For instance, coherent-QPSK detection has an
OSNR margin of .about.2.4 dB at a BER of 10.sup.-3 with respect to
differential-QPSK.
SUMMARY OF THE INVENTION
[0016] It is an object of the present invention to provide improved
apparatus and methods. According to a first aspect of the present
invention, there is provided an optical receiver arranged to
receive and demodulate optical orthogonal frequency division
multiplexed signals, and having
[0017] a subcarrier reference generator arranged to generate a
number of subcarrier reference signals, each for use in
demodulating a different one of a number of frequency channels of
the frequency division multiplexed signals,
[0018] the subcarrier reference generator further being arranged to
compensate for degradations in the generated reference signals by
averaging a number of estimates derived from different inputs.
[0019] By using an average of a number of estimates, the reference
can be made more accurate, and so more resilient to any of a
variety of degradations. Embodiments of this invention can thus
complement OFDM by enabling longer system reach and/or increased
optical power margins by means of better system resilience to
amplification noise and reduced source coherence. If the reference
is more accurate, the bit error rate can be reduced, and/or the
capacity can be increased by increasing bit rate or introducing
more frequency channels. The estimates can encompass estimates of
the reference signal values or estimates of the values of the
degradation for example. The different inputs can encompass inputs
such as data received at different time instances or of different
frequency channels or partially demodulated or partially corrected
data or combinations of these and other parameters for example. The
average can be a weighted average, weighted to favor inputs closer
to the reference value concerned, or can be any type of combination
of the inputs for example.
[0020] An additional feature of some embodiments, suitable for a
dependent claim is, the generator being arranged to compensate for
phase drift by determining estimates at a number of different
frequencies and averaging these estimates. Phase drift errors can
be a particular limitation for optical systems since laser sources
typically introduce considerable phase drift over the relevant time
intervals. For a given transmission rate, the more channels there
are, the greater the time between successive data portions of the
same channel, and so the greater is the susceptibility to phase
drift. Averaging over a number of frequencies can enable a more
accurate determination of the drift and therefore more accurate
compensation for it.
[0021] Another such additional feature is the generator being
arranged to compensate for noise by determining estimates for a
given one of the reference signals at a number of different times
and averaging these estimates. Such a time averaging can enable the
reference to be less susceptible to noise such as Gaussian
noise.
[0022] Another such additional feature is the generator being
arranged to compensate for phase drift by determining estimates at
a number of different frequencies and averaging these estimates,
and to compensate for noise by determining estimates for a given
one of the reference signals at a number of different times and
averaging these estimates. This combination can have some
synergistic effect, in that noise performance can be improved if
the averaging is carried out on phase drift corrected estimates and
vice versa.
[0023] Another such additional feature is the optical receiver
being arranged to demodulate differentially coded optical
orthogonal frequency division multiplexed signals, the generator
being arranged to operate without using a transmitted pilot tone.
The use of differentially coded signals is useful as it does not
require the transmission of a pilot tone for demodulation.
[0024] Another such additional feature is the optical receiver
being arranged to demodulate non differentially coded optical
orthogonal frequency division multiplexed signals. This can give
improved phase noise resilience with respect to current common
phase noise reduction techniques.
[0025] Another such additional feature is the generator being
arranged to generate the estimated reference signals by stripping
detected data from a received signal for each channel. This is a
key part of a decision directed algorithm and is an efficient way
of generating the reference signals because it minimizes the impact
of additive Gaussian noise in the estimation of the reference.
[0026] Another such additional feature is the generator being
arranged to generate an estimated phase drift by estimating a
rotation of a constellation of points represented in a frequency
domain. In this technique, for each received symbol a constellation
is built that includes all the unprocessed subcarriers detected.
This constellation is subsequently rotated for a different number
angles and the overlap with the ideal constellation is evaluated by
means of a cost function. The angle yielding a minimum for this
cost function represents an estimate of the phase drift.
[0027] This is an alternative way to estimate the phase drift that
is very superior for formats such as 16-32 QAM, where phase drift
estimation with data stripping algorithms would not be applicable
due to their low tolerance to phase drifts. This lack of phase
tolerance is due to the tightly packed constellation exhibited by
this modulation formats.
[0028] Another such additional feature is the optical receiver
being arranged to receive signals carrying data bits which have
been encoded before transmission by a rotation in a complex
frequency domain, relative to other data bits and the receiver
having a decoder for decoding by using the other data bits to
determine an amount of rotation of the data bits, needed to decode
the data bits. An advantage of this is that such relative encoding
enables more resilience to phase jumps of greater than 45 degrees
between symbols, which become more likely with longer symbols.
[0029] Another such additional feature is a coherent optical domain
receiving section, and an electrical demodulation section arranged
to carry out a Fast Fourier Transform, and to use the subcarrier
reference signals for detection of data in the fourier transformed
signals. This is a particularly efficient way of dividing the
processing between optical and electrical domains, and of achieving
demodulation, other ways are conceivable in principle.
[0030] Another aspect of the invention provides a demodulator for a
receiver for orthogonal frequency division multiplexed signals, and
having
[0031] a subcarrier reference generator arranged to generate a
number of subcarrier reference signals, each for use in
demodulating a different one of a number of frequency channels of
the frequency division multiplexed signals,
[0032] the subcarrier reference generator further being arranged to
compensate for degradations in the generated reference signals by
averaging a number of estimates derived from different inputs.
[0033] Another additional feature is the generator being arranged
to compensate for phase drift by determining estimates at a number
of different frequencies and averaging these estimates, and to
compensate for noise by determining estimates for a given one of
the reference signals at a number of different times and averaging
these estimates.
[0034] Another such additional feature is the demodulator being
arranged to demodulate differentially coded optical orthogonal
frequency division multiplexed signals, the generator being
arranged to operate without using a transmitted pilot tone.
[0035] Another such additional feature is the demodulator being in
the form of software. This acknowledges that software can be a
valuable, separately tradable commodity. It is intended to
encompass software, which runs on or controls "dumb" or standard
hardware, to carry out the desired functions, (and therefore the
software essentially defines the functions of the demodulator, and
can therefore be termed a demodulator, even before it is combined
with its standard hardware). For similar reasons, it is also
intended to encompass software which "describes" or defines the
configuration of hardware, such as HDL (hardware description
language) software, as is used for designing silicon chips, or for
configuring universal programmable chips, to carry out desired
functions.
[0036] Another aspect provides a subcarrier reference generator for
a demodulator for orthogonal frequency division multiplexed
signals, the generator being arranged to generate a number of
subcarrier reference signals, each for use in demodulating a
different one of a number of frequency channels of the frequency
division multiplexed signals,
[0037] the subcarrier reference generator further being arranged to
compensate for degradations in the generated reference signals by
averaging a number of estimates derived from different inputs.
[0038] Another aspect of the invention provides a subcarrier
reference generator for generating subcarrier reference signals for
use in demodulating optically transmitted orthogonal frequency
division multiplexed signals, by generating a number of subcarrier
reference signals, each for use in demodulating a different one of
a number of frequency channels of the frequency division
multiplexed signals,
[0039] and comprising means for compensating for degradations in
the generated reference signals by averaging a number of estimates
derived from different inputs.
[0040] This corresponds to the first aspect but uses "means for"
language, to give different claim scope.
[0041] Another aspect provides an optical receiver arranged to
receive and demodulate optical orthogonal frequency division
multiplexed signals carrying higher order QAM data values encoded
by a mapping in a complex frequency domain according to
corresponding other data values, the receiver having
[0042] a subcarrier reference generator arranged to generate a
number of subcarrier reference signals,
[0043] a demodulator for demodulating the optical orthogonal
frequency division multiplexed signals using each of the subcarrier
reference signals for a respective frequency channel of the
frequency division multiplexed signals, and
[0044] a decoder for decoding after the demodulating, by
determining from the corresponding other data values, an inverse
mapping in the complex frequency domain needed to decode the data
values.
[0045] An advantage of such relative encoding and relative
decoding, using other data values rather than an absolute
reference, is that it can enable more resilience to phase jumps of
greater than + or -45 degrees between symbols, which become more
likely with longer symbols. Such advantages can apply with or
without the subcarrier generator with averaging discussed
above.
[0046] As an additional feature, the mapping and inverse mapping
comprises a rotation.
[0047] An advantage of using rotational encoding is that it can
provide better resilience to such phase jumps than other
mappings.
[0048] As an additional feature, the rotation can be any of 0, 90,
180, or 270 degrees. This is one way of allowing rotation of a
conventional QAM constellation of points without altering the
constellation. The constellation can be divided into groups of
points forming a triangle or a square for example. Other angles or
ways of grouping can be conceived.
[0049] Another aspect of the invention provides an optical
transmitter arranged to transmit an optical orthogonal frequency
division multiplexed signal carrying higher order QAM data values,
the transmitter having:
[0050] an encoder for encoding the data values by carrying out a
mapping in a complex frequency domain according to corresponding
others of the data values, and
[0051] a modulator for modulating the encoded data values to form
the optical orthogonal frequency division multiplexed signal having
a number of frequency channels.
[0052] Another aspect provides a method of generating subcarrier
reference signals for use in demodulating optically transmitted
orthogonal frequency division multiplexed signals, by generating a
number of subcarrier reference signals, each for use in
demodulating a different one of a number of frequency channels of
the frequency division multiplexed signals,
[0053] and by compensating for degradations in the generated
reference signals by averaging a number of estimates derived from
different inputs.
[0054] Another such additional feature is the step of compensating
for phase drift by determining estimates at a number of different
frequencies and averaging these estimates.
[0055] Another aspect provides a method of offering a data
transmission service over an optical link, the data being
transmitted using optical orthogonal frequency division multiplexed
signals, and being received using an optical receiver having a
subcarrier reference generator arranged to generate a number of
subcarrier reference signals, each for use in demodulating a
different one of a number of frequency channels of the frequency
division multiplexed signals, the subcarrier reference generator
further being arranged to compensate for degradations in the
generated reference signals by averaging a number of estimates
derived from different inputs. This reflects the position that if
the link is enhanced, then data transmission services over the
network can be enhanced, and the value of such services can
increase as a direct consequence. Such increased value of
transmission services over the life of the system, could prove far
greater than the sales value of the equipment.
[0056] Any of the features can be combined with each other and with
any of the aspects of the invention as would be apparent to those
skilled in the art. Other advantages will be apparent to those
skilled in the art.
BRIEF DESCRIPTION OF THE DRAWINGS
[0057] To show by way of example how the invention can be
implemented, embodiments will now be described with reference to
the figures in which:
[0058] FIG. 1 shows an example of an optical receiver according to
an embodiment,
[0059] FIG. 2 shows an example of a subcarrier reference generator
which can be used in the receiver of FIG. 1, and uses time
averaging, according to an embodiment,
[0060] FIG. 3 shows a graphical representation of the characters of
the data symbols,
[0061] FIG. 4 shows an example of phase drift correction for a
subcarrier reference generator according to another embodiment,
using frequency averaged phase drift correction,
[0062] FIG. 5 shows a graphical representation of the characters of
the symbols in the frequency domain, showing the effect of
frequency averaging,
[0063] FIG. 6 shows another example of phase drift correction using
constellation rotation,
[0064] FIG. 7 shows an example of a transmitter having rotational
encoding,
[0065] FIGS. 8 and 9 show such encoding and corresponding decoding
at the receiver in schematic form, and
[0066] FIG. 10 shows a subcarrier reference generator with time
averaging and frequency averaged phase drift correction according
to another embodiment.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
[0067] Some embodiments described address how to enhance the noise
robustness of OFDM without the overhead of transmitting pilot
tones. The technique described can be called decision driven
averaged carrier recovery. Other embodiments introduce a method to
correct for phase drifts in the optical source also without need of
pilot tone transmission. Embodiments of the demodulation techniques
described can approach the noise resilience of the coherent limit
but without the additional overhead of the pilot tones. The data
can be differentially encoded in the transmitter, so that
transmission of any reference signals or pilot tones is not always
necessary. The combination of both techniques can be used to
minimise the impact of the amplification noise and the reduced
laser coherence that would otherwise affect optical OFDM.
[0068] Minimising the impact of additive white Gaussian noise
involves generating a subcarrier reference also termed a Reference
Subcarrier Complex-Amplitude (RSCA) for each of the frequency
channels by means of a decision directed algorithm. For each
channel, the first symbol is taken as initial RSCA, which will be
upgraded symbol after symbol. Each new sampled symbol is compared
with respect to the current RSCA and, as a result of the
comparison, a decision is made about its value. One step in this is
to strip the estimated discrete value from the actual complex value
of the noisy received symbol to obtain a new estimate of the
reference subcarrier RSCA. The new estimate of RSCA is used in
conjunction with the previous reference to improve its accuracy by
means of weighted average between them.
[0069] The correction of the common phase errors involves averaging
over the different subcarriers at a given moment in time. In this
patent we describe two possible alternatives. The first one is
based on data stripping and is particularly useful for modulation
formats with low complexity, like QPSK and its differential
versions. The second technique is based on estimation of optimum
constellation rotation for a given symbol, and it exhibits clear
superiority for more complex modulation formats like 16-32 QAM.
[0070] The first algorithm is illustrated for the QPSK modulation
format and proceeds in two stages. First for each frequency channel
(subcarrier) the angle difference between the received symbol's
complex amplitude and expected RSCA is calculated. This angle is
then multiplied by four (mod 2.pi.) to strip the data. This is
equivalently to data stripping by a fourth order phase lock loop.
The result will give an idea of the phase drift of the source in
the time scale corresponding to the whole 32-symbol length. In a
second stage the angles previously calculated for each of the
frequency channels are averaged. The result is divided by four to
estimate a unique phase drift value. This phase drift value is
applied to each of the RSCAs to estimate the phase drifted RSCA
before the data demodulation. In summary, the phase noise
correction is obtained by looking into the correlations among the
various frequency channels.
[0071] In the second technique, phase drift is estimated by
evaluating the rotation of the constellation of points represented
in the frequency domain. For each received symbol a constellation
is built that includes all the unprocessed subcarriers detected.
This constellation is subsequently rotated for a different number
angles and the overlap with the ideal constellation is evaluated by
means of a cost function. The angle yielding a minimum for this
cost function represents an estimate of the phase drift. This is an
alternative way to estimate the phase drift that is useful for
formats such as 16-32 QAM, where phase drift estimation with data
stripping algorithms would not be applicable due to their low
tolerance to phase drifts. This lack of phase tolerance is due to
the tightly packed constellation exhibited by this modulation
formats.
[0072] The applicability of these techniques for subcarrier
recovery, phase error correction, and demodulation extends to any
possible phase/amplitude modulation format used for coding the OFDM
symbols. This will include BPSK, QPSK, 16-32 QAM and similar
modulation formats. The techniques can address how to enhance AWG
noise robustness and combat common phase noise in OFDM without the
overhead of transmitting pilot tones.
[0073] FIG. 1 Optical Receiver for Receiving OFDM Signals.
[0074] FIG. 1 shows some of the principal elements of an optical
receiver for receiving optical orthogonal frequency division
multiplexed signals. It includes an optical front end for
coherently receiving the optical signal from the optical
transmission line. This part includes an optical mixer 100, and a
local oscillator 140, and output IQ signals. These are converted to
electrical form by differential optical to electrical converters
110. These output analogue signals which are passed to analogue to
digital converters 120. Here they are converted to digital signals
by sampling with a sampling clock 130. At this point, the digital
signals represent time domain complex amplitude sequences. The
functions after this point can be carried out by circuitry such as
digital signal processing hardware, microprocessors, or application
specific integrated circuits, as appropriate. The functions can be
implemented in conventional software languages to run on the
hardware. To achieve real time processing at data rates
conventionally used in existing optical systems, fast hardware is
needed, though techniques such as parallel processing can be used
to enable clock rates to be lowered.
[0075] At 150 the sequences are grouped into the symbols that were
transmitted. In a typical example there may be 32 data characters
in the form of one or more complex numbers depending on the
modulation format, making up each symbol. Guard bands are
conventionally provided between the symbols, to allow for channel
degradation. These are removed by this function. Next a fast
Fourrier transform (FFT) 160 function is carried out on each
symbol. This returns each symbol to the frequency domain, thus the
digital samples represent frequency domain complex amplitude
sequences. As shown in the graph included in FIG. 1, at this point
the data can be represented graphically with the x axis
representing time, and the y axis representing different
frequencies. A first symbol includes the boxes illustrated as being
at the same time instant, but at different frequencies. This is
illustrated by the first column of boxes, where each box represents
one data character, which may be a complex value, or for some
modulation formats, eg 16 QAM, (Quadrature Amplitude Modulation) a
constellation of 16 complex values.
[0076] As represented by the graph in FIG. 1, after the FFT
operation, each frequency channel is represented by a horizontal
row of boxes at the same frequency for successive time instances.
At function 180, the data characters for each channel are subject
to amplitude and phase noise, and are typically multi-bit digital
values, which need to be compared to a subcarrier reference for
that channel, to decide a true discrete logical value for each data
character. Hence subcarrier reference generators 190 are provided
for each channel. At function 200, if the data is transmitted as
differential data, logic is provided to convert differential
characters to the original data, and to carry out FEC (forward
error correction) processing. The detection stage represented by
functions 180 and 200 can incorporate techniques such as MLSE as
described in more detail in the above referenced pending
application Ser. No. 10/626,384.
[0077] The operation of these receiver functions will now be
described. An incoming OFDM-Amplitude/Phase modulated optical
signal is coherently detected by optically mixing with the local
oscillator. This is usually achieved by optical hybrids and
differential detectors. Sampling the output of these detectors
produces the time domain in-phase and quadrature components. The
standard OFDM demodulation involves grouping the received sequence
in blocks, removing the guard-bands for each block, and carrying
out a fast Fourier transform to recover the frequency-domain
transmitted complex amplitude sequence for one symbol after
another. For each frequency channel (subcarrier), a sequence is
recovered with the original amplitude/phase modulation format. The
complex symbol sequence for each frequency channel appears
multiplied by the spectral response of the channel corresponding to
that frequency, and is also corrupted by noise.
[0078] As discussed previously, in order to demodulate the data a
differential comparison between adjacent complex symbols can be
carried out, with the corresponding noise penalty; or recover a
reference subcarrier from transmitted pilot tones, with the
corresponding capacity overhead. The alternative solution described
here is to generate the subcarrier with the help of a decision
directed averaging algorithm in order to obtain almost the same
noise resilience of the pilot-tone approach but without its
capacity overhead. Optionally this can also involve a common phase
error correction as part of the process, described below with
reference to FIGS. 4-7. The technique developed to minimise the
impact of additive white Gaussian noise is called decision driven
averaged carrier generation for each of the frequency channels. It
can be carried out by the channel subcarrier reference generator
function 190. This can generate the reference signals according to
an average of a number of estimates. This part will be described in
more detail with reference to FIG. 2 onwards.
[0079] FIGS. 2, 3 Subcarrier Generator Using Time Averaging
[0080] FIG. 2 shows some of the principal elements of a subcarrier
generator which exploits time averaging to overcome degradations
due to Gaussian noise. At step 320 a raw noisy symbol before
decision by part 180 of FIG. 1's input. The discrete value after
the decision is stripped from the raw complex value, to give a new
estimate of the subcarrier reference (RSCA). Function 300 shows
obtaining an initial RSCA from an initial raw symbol for the
initial state that no decision has yet been made. At function 330 a
weighted average of the new and prior RSCA values is obtained.
There are a number of ways of implementing this. For example, a
number of prior values can be stored in a shift register, and the
parallel outputs of the shift register can be weighted and added.
Alternatively a single prior value can be stored, and a proportion
of the new value added to the prior value. The result of the
addition can be used at the time averaged output, and at the same
time can be fed back to be stored as the prior value for the next
time interval. Function 340 shows how the time averaged RSCA is
used in the comparison with the raw noisy symbol, to decide the
discrete value. This is essentially equivalent to step 180 of FIG.
1. A subcarrier reference is generated for each of the subcarriers,
and a decision is made separately for each of the subcarriers. For
a given time interval, the decisions of each of the characters
using the different subcarriers are carried out, and the individual
characters can be reassembled into the symbol transmitted. The next
symbol can then be assembled in the same fashion.
[0081] FIG. 3 is a representation of the effect of the time
averaging carried out by the elements of FIG. 2. As in the
graphical representation in FIG. 1, the symbols are represented by
columns, and each subcarrier is represented by rows. Each box
represents a character, which is a complex value or constellation
of complex values. The horizontal arrows along the rows represent
the effect of time averaging. Individual estimates of the
subcarrier reference are obtained for each of the boxes. The effect
of the time averaging is to take an average of a number of the
boxes along a single row.
[0082] In other words, for a particular frequency channel, the
subcarrier recovery algorithm starts by taking the first noisy
amplitude as initial Reference Subcarrier Complex-Amplitude (RSCA).
The second noisy symbol is then processed by comparing it first
with the RSCA, and making a decision about its value (relative to
the reference). Once that decision is made, the data on the second
noisy symbol is stripped to align it approximately with the initial
reference. A new reference signal is then calculated as a weighted
average of the current reference estimate and the initial
reference.
[0083] The process then carries on iteratively. A new noisy complex
symbol is considered; its value is estimated against the current
averaged RSCA; this value is subsequently stripped from the noisy
symbol, which is left approximately aligned with the current
reference signal. Finally a new estimation of the RSCA is made.
[0084] The weight attached to the present estimate of the reference
signal in the averaging process governs the effective number of
symbols over which the averaging is carried out. The smaller this
number, the larger the number of symbols involved in the averaging.
In order to approach the coherent BER-OSNR limit a value of 0.25 or
lower can be used. However, even with a value of 0.5 a noticeable
benefit in noise resilience can be obtained, for a given size of
symbol and other parameters.
[0085] The BER-OSNR characteristics can be measured for optical
OFDM with various parameters. One example uses QPSK formatting (10
Gbs, 32 subcarriers, so 32 characters per symbol), other examples
can use differential-QPSK. Better performance in terms of BER-OSNR
characteristic can be obtained for smaller weighting factors, i.e.
more averaging, and can reach the coherent limit.
[0086] FIG. 4-6 Subcarrier Reference Generation with Phase Drift
Correction
[0087] FIG. 4 shows in schematic form some of the principal
elements of subcarrier reference generation with phase drift
correction. At function 500, an angle difference is calculated
between the current subcarrier reference RSCA and the received
complex symbol in its raw state before decision. Next, the discrete
data value is stripped from this difference, for each subcarrier to
obtain an estimate for the phase drift over the symbol period, at
510. Next, at 520 an average of these angles for some or all of the
frequency channels for the given symbol is obtained. This results
in a frequency averaged phase drift value. At 540, the frequency
averaged phase drift is applied to correct the subcarrier reference
RSCA of each of the channels.
[0088] This process is represented graphically in FIG. 5. This
figure corresponds closely to FIG. 3. In this case, the arrows
extending vertically along the columns of boxes represent the
effect of frequency averaging of the phase drift correction value.
The phase drift is estimated for each of the boxes. Each box
represents a data character of a symbol. The frequency averaged
phase drift value is determined by averaging the phase drift values
calculated for each of the boxes in a column as shown in FIG.
5.
[0089] In other words, the algorithm proceeds in two stages. First
the phase drift is estimated. For low complexity modulation formats
like QPSK for each frequency channel (subcarrier) the angle
difference between the received symbol's complex amplitude and
expected RSCA is calculated, the data are then stripped, and the
residual angle differences averaged. For more complex modulation
formats, like 16-32 QAM, the constellation rotation technique of
FIG. 6 would be used and the optimum rotation angle estimated. The
result represents the phase drift of the source in the time scale
corresponding to the whole 32 character length of the symbol. In a
second stage the angles previously calculated for each of the
frequency channels are averaged, and the result is divided by four
to estimate a unique phase drift value. This phase drift value is
applied to each of the RSCA to estimate the phase drifted RSCA
before the data demodulation. In summary, the phase noise
correction is obtained by looking into the correlations among the
various frequency channels.
[0090] This frequency averaging enables the impact of phase noise
to be reduced and a more accurate correction for common phase drift
among all the subcarriers to be made. This is particularly
important given the reduced coherence of optical sources.
[0091] FIG. 6 shows a constellation rotation technique for
estimating the phase drift by evaluating the rotation of the
constellation of points represented in the frequency domain. For
each received symbol a constellation is built at step 600. This
includes all the unprocessed subcarriers detected. This
constellation is subsequently rotated at step 610 for a number of
different angles. At each angle, there is an evaluation of the
alignment with an optimum rotation angle using a cost function. At
step 620 the angle yielding a minimum for this cost function is
determined. This represents an estimate of the phase drift. As
before, at step 540 the average phase drift is used to correct the
RSCA of each channel.
[0092] Such a constellation-directed algorithm can track the
reference phase very effectively to address the issue of laser
phase noise which can be a major source of degradation in OFDM
optical systems due to reduced coherence of optical sources and the
long duration of symbols. Such algorithms could suffer from an
occasional cycle slip if the common phase jump for a symbol is
higher than (+/-.pi./4). Such phase-induced cycle slips in high
order QAM-OFDM can have catastrophic effects because the error will
propagate. The problem does not apply to differentially coded
techniques such as examples using QPSK as in FIG. 4, since this
does not rely on an absolute phase reference. One way to address
this problem for the examples using higher order e.g. 16 or 32 QAM
as in FIG. 6 will now be discussed. It involves incorporating
partially-differential encoding at the transmitting side and a
corresponding decoding at the receiving side.
[0093] FIGS. 7-9
[0094] FIG. 7 shows an example of a transmitter incorporating the
encoding, for use with a corresponding receiver having a decoder.
FIG. 8 shows schematically the relative rotation encoding carried
out in the transmitter of FIG. 7. FIG. 9 shows schematically the
corresponding relative rotation for the decoding in the receiver.
In FIG. 7, input data in the form of a QAM constellation at step
710 is encoded by a relative mapping such as a rotation, as will be
explained in more detail below. At step 720, an IFFT operation is
carried out to convert frequency domain complex amplitude data to
the time domain. Guard bands are added at step 730. The transmitter
also includes D-A convertors 740 clocked by clock 750. Electronic
drivers 760 feed an I-Q modulator 770 to modulate light from a
laser 780 to create the optical signal for transmission.
[0095] The example of the encoding and decoding algorithm described
here involves dividing the (QAM) constellation in four quadrant
groups of points as shown in FIGS. 8 and 9. Each quadrant can be
assigned a QPSK value (from 1 to 4), and the points of the
constellation in each quadrant can be called a QPSK-group. The
algorithm involves encoding the data value for each subcarrier in
symbol n by reference to the value of the same subcarrier in symbol
n-1. If it is assumed that the information in a particular
subcarrier is given by the string of M-QAM {x.sub.n} where n=1 to
N. The encoding process produces the set of values {y.sub.n} from
where n=0 to N.
[0096] The encoding process for x.sub.n is shown in FIG. 8. It has
the following 2 steps between the 2 pairs of states shown in the
figure. Both steps should be carried out in the period of one data
value:
[0097] 1. Take y.sub.n-1 known from previous encoding, (or
initially it is set as y.sub.0 arbitrarily), then identify the
QPSK-group to which y.sub.n-1 belongs, or rotate the reference
frame until y.sub.n-1 belongs to Group 1. The angle of rotation
.phi. can have the values 0, .pi./2, .pi., 3.pi./2.
[0098] 2. Take the constellation point corresponding to x.sub.n in
the rotated frame. Rotate back x.sub.n by -.phi. found in step 1,
to obtain the encoded y.sub.n, then make n=n+1 and repeat
steps.
[0099] The result of the encoding is data values which are
partially differential relative to the preceding data value. They
are complex values with a reference frame whose rotation in steps
of 90 degrees which is set by the preceding data value. But each
value is not expressed as a difference from the preceding data
value, so they are not differential values. Decoding is shown
schematically in FIG. 9. In the receiving side, decoding is carried
out after the FFT and comparison of each frequency channel to its
subcarrier reference, so can be carried out by element 200 of FIG.
1. Decoding can be represented as converting a received set
{z.sub.n} back to {x.sub.n}. It is assumed that the received set
{z.sub.n} is equal to the original transmitted encoded set
{y.sub.n} except for sudden cycle slips. Cycle slips are rotations
of +/-.pi./2. If the cycle slip operators are represented by
C.sup.+/-, a typical received sequence {z.sub.n} will be:
[0100] {y.sub.0, y.sub.1, y.sub.2, y.sub.3, C.sup.+y.sub.5,
C.sup.+y.sub.6, C.sup.+C.sup.+y.sub.7, C.sup.+C.sup.+y.sub.8,
C.sup.+C.sup.+y.sub.9, C.sup.+C.sup.+y.sub.10,
C.sup.-C.sup.+C.sup.+y.sub- .11, C.sup.-C.sup.+C.sup.+y.sub.12,
C.sup.-C.sup.+C.sup.+y.sub.13, C.sup.-C.sup.+C.sup.+y.sub.14,
C.sup.+C.sup.-C.sup.+C.sup.+y.sub.15 . . . }
[0101] where the cycle slip events are underlined.
[0102] The decoding algorithm can be shown as 2 steps between the 2
pairs of states shown in FIG. 9:
[0103] 1. Take z.sub.n-1 & z.sub.n and find the angle .psi.
required for z.sub.n-1 to be rotated to the first QPSK group. The
angle of rotation .psi. can have the values 0, .pi./2, .pi.,
3.pi./2, depending on which quadrant or QPSK-group y.sub.n-1
belonged to.
[0104] 2. Take z.sub.n and rotate it by this angle .psi. to give:
T.sub.104 z.sub.n. Take this as the decoded value x.sub.n equal to
T.sub..psi.z.sub.n. Make n=n+1 and repeat steps.
[0105] The decoded data will be in error if there is a cycle slip,
but the error will not propagate to following symbols due to the
relative encoding. Due to the averaging techniques the degradation
in required OSNR should be minimal.
[0106] FIG. 10 Subcarrier Reference Generator with Combination of
Time Averaging and Frequency Averaged Phase Drift Correction.
[0107] FIG. 10 shows in schematic form some of the principal
functions in a subcarrier reference generator 190 with combination
of time averaging and frequency averaged phase drift correction. A
first step in the subcarrier reference generator is to estimate the
phase drift error and obtain a frequency averaged error for each
symbol, at step 725. The averaged error determined for a given
symbol is used to correct the subcarrier references for all the
channels at the time of that symbol. The phase drift corrected
subcarrier references are then used by function 735 to obtain
weighted time averages of new and prior subcarrier reference values
for each subcarrier. As shown by function 745, these accurate
subcarrier references are used for the comparison with the raw
noisy values of the characters of each symbol to decide the
discrete values of each character, shown by function 745. The
corrected subcarrier references are fed back within the subcarrier
reference generator for use in the various functions.
[0108] This combination can simultaneously combat additive white
Gaussian noise and phase-noise. The amount of benefit will depend
on the type of system and its parameters. The phase noise of a
laser is usually measured by its spectral linewidth. Assuming that
this is lorentzian, the beneficial combined effects of the decision
directed averaging algorithm and the phase correction algorithm
working together can be considerable for linewidths of the order of
200 kHz in systems of about 5 GSamples/s. The phase noise error
correction algorithm on its own can reduce a penalty of 1 dB in
OSNR to about 0.25 dB. For this linewidth the decision directed
averaging process in conjunction with the phase equalisation
scheme, with averaging over two samples, (approx equivalent to
AV=0.5) an improvement of 1 dB of OSNR was seen over the standard
performance of DQPSK (with an ideal laser). A loss of only
.about.0.4 dB was seen with respect to the averaged-DQPSK with the
same AV(=0.5). The OSNR for a BER of 10.sup.-3 for 10
Gbs-OFDM-DQPSK can be close to 4.3 dB/0.1 nm for a linewidth of 200
kHz. Both phase noise equalization and decision directed averaging
are used to achieve this value. This is .about.1 dB better than
standard DQPSK.
[0109] Concluding Remarks
[0110] Although the reference generators have been illustrated for
the case where pilot tones are not used, in principle these
techniques can be combined with the use of a pilot tone either to
generate the subcarrier references, or to verify or correct
references that have been generated from the raw symbols.
[0111] Embodiments of this invention can enable improvements in
Optical OFDM, to provide a highly flexible and scalable optical
communication system. Optical OFDM can show resilience to high
levels of linear distortion (chromatic and polarisation mode
dispersion). This will enable transmission over highly dispersive
channels without the need of costly optical compensators.
Embodiments of this invention can complement OFDM by enabling high
reach and/or increased margins by means of better system resilience
to amplification noise and reduced source coherence. Increasing the
noise resilience of the communication system has important
commercial implications in terms of the final cost of the
transmission system. For instance, improving the noise resilience
by a factor of 1.6 (2 dB) will enable transmission line operators
to reduce their capital and running cost expenditures in optical
amplifiers by the same factor. In summary, OFDM cost savings can
now relate not only to reduction or removal of linear distortion
compensators in the transmission line, but also to a reduced number
of required amplifiers as a result of being able to exploit the
coherent detection limit without capacity penalties due to
transmitted overheads.
[0112] It can also be applied to generating subcarrier references
for non optical OFDM receivers such as radio systems, particularly
in cases where phase noise or phase drift is an issue.
[0113] As has been described above, an optical receiver demodulates
optical orthogonal frequency division multiplexed signals and
generates a number of subcarrier reference signals, each for
demodulating a frequency channel of the frequency division
multiplexed signals. It compensates for degradations in the
generated reference signals by averaging a number of estimates
derived from different inputs to make the references more resilient
to degradations. It can encompass time averaging to compensate for
amplification noise, and frequency averaging of phase drift
estimation to compensate for phase drift caused by reduced source
coherence. It can enable longer system reach and/or increased
optical power margins by means of better system resilience to
amplification noise and reduced source coherence. The bit error
rate can be reduced, and/or the capacity can be increased by
increasing bit rate or introducing more frequency channels.
[0114] Other variations will be apparent to those skilled in the
art, having corresponding advantages to those set out above, within
the scope of the claims.
* * * * *